Can SDN Produce a Single Data Network?

One of the more interesting aspects of NFV, SDN and other forms of virtualized networking is the way they break down the barriers that exist within physical network infrastructure.

Currently, the data center network and both the local and wide area networks consist of different topologies and play unique roles in the data connectivity chain. While that will continue as a factor in network deployment going forward, operationally the lines are blurring under increasingly complex layers of abstraction.

For the enterprise, this means greater flexibility when it comes to extending data center connectivity across great distances, but it can also lead to a more convoluted management stack in which the traditional metrics of network performance may or may not apply in any given situation.

Companies like Aerohive are at the forefront of breaking down the barriers that exist between physical networks. The company recently launched an SD-LAN portfolio that seeks to bring the same software-defined abstraction that is infiltrating the data center network and, increasingly, the WAN, to the local area. The portfolio includes a new line of cloud-managed access switches and a 5 Ghz wireless access point, all linked under the HiveManager NG cloud-based policy engine that utilizes open APIs to drive dynamic application and identity-driven network performance. In this way, the company hopes to drive the operational efficiency and scalability of SDN across the corporate office and into the campus network.

Crafting a common management platform across multiple network topologies is one of the prime objectives of network virtualization, according to SDX Central’s Michael Vizard, but it is unclear how easy this will be given the plethora of competing solutions entering the channel. On the one hand, some enterprises are exploring the possibilities around linking seemingly disparate platforms like VMware’s NSX with Cisco’s ACI portfolio, but then there are innovative new approaches to the SD-WAN from start-ups like Viptela and VeloCloud, not to mention emerging partnerships like the recent effort to implement Avaya’s SDN Fx management stack on FatPipe appliances. It is unlikely that a single enterprise will have to contend with all of these solutions – although anything is possible in a distributed cloud environment – but at the very least network managers will have to determine whether their needs are best served under a single-vendor or vendor-alliance solution or a standards-based virtual network architecture.

Despite the smooth integration between the SD-LAN and SD-WAN, it is important to remember that these topologies still maintain some important distinctions, says Viptela’s Ramesh Prabagaran. The LAN, after all, is predictive in nature, with multiple switches and interconnects designed to aggregate bandwidth. The WAN uses hub-and-spoke and mesh topologies, as well as regional aggregation and traffic-engineered pathways that sit atop multiple carrier networks. In addition, there are differences in security, performance characteristics and a host of other factors that affect how they are deployed and orchestrated. Again, abstraction can mask these differences behind unified management and automation stacks, but don’t expect to one day find yourself working with an end-to-end, one-size-fits-all SDN architecture.

Complicating matters further is the fact that the enterprise is facing an all-new network environment to supplement its data center/LAN/WAN infrastructure. Big Data and the Internet of Things will require unique networking characteristics of their own, including dramatically more bandwidth and extreme degrees of flexibility. This will lead to what IT Pro Portal calls The Third Network (the enterprise and public carrier networks being One and Two), which combines elements of Carrier Ethernet 2.0 with public Network as a Service platforms to provide on-demand connectivity, security and stability in the face of extremely demanding workloads. A key component of the Third Network is the Lifecycle Service Orchestration API that defines an end-to-end connectivity standard that can support IoT applications across wired and wireless infrastructure.

Clearly, the life of a network engineer in a software-defined data ecosystem will be full of challenges, but as long as the virtual network layer is architected properly the primary task will be to develop the appropriate solutions and let the automated management stack do the tedious work of actually implementing them.

Abstraction may one day give the appearance of a single, integrated entity, but in reality the traditional network layers will continue to do their specific jobs as they always have.


Latest Articles

Follow Us On Social Media

Explore More